Vinnumarkaður - 01.06.1995, Blaðsíða 26

24
Synopsis of methods and concepts
3. Synopsis of Methods and Concepts
The main part of this synopsis deals with the Icelandic
Labour Force Survey (ILFS), its methods, reliability and
principal concepts. Other data sources are described in a
similar manner.
3.1 The Icelandic Labour Force Survey
3.1.1. Previous labour force statistics
The principal source for labour force statistics in Iceland has
for a long time been based on information on working weeks
derived from tax retums as well as the official registration
of unemployment. These data, however, have many short-
comings, e.g., little possibility of analysis according to
various socio-economic groups, lack of important variables
such as working hours and status in employment, and lack
of intemational comparability.
For these reasons Statistics Iceland (SI) decided to start its
own labour force survey in 1991 based on international
standards. The first such survey was conducted in April
1991 and since then there has been a survey twice a year, in
April and November.
3.1.2. Execution
Questionnaire. The ILFS questionnaire is based on compa-
rable questionnaires in the Nordic countries. The question-
naire has also been developed so that standardized data can
be sent to Eurostat according to the EEA agreement. Ques-
tions of particular national interest have also been included.
In the course of the first three years of the ILFS the
questionnaire has changed considerably, even though the
main core of questions has remained the same. Many
questions have been added, some have been discarded while
others have been rephrased.
Survey periods and number of interviewers. Each
survey has one reference week. As each survey is conducted
for 10-11 days, the reference week is moved forward after
the seventh day for the remaining respondents (Table 2.1).
Usually the first reference week is the first or second week
of the interview month.
From November 1991 interviewers have been hired from
outside the SI. For the most part the interviewers have had
previous experience in working with telephone surveys.
Each survey is conducted by approximately 20 interviewers.
Before each survey the interviewers are trained in conduct-
ing interviews and working with the computer aided inter-
viewing system.
Data gathering. All interviews in the ILFS are telephone
interviews. Most interviews take place in the evenings and
during weekends, except for interviews by appointment
during normal working hours. Telephone numbers for re-
spondents are provided by the Post and Telecommunication
Administration. Considerable effort is made to reach par-
ticipants who have moved or do not have a telephone. No
participant is visited, however.
In November 1992 computer aided telephone interview-
ing (CATI) was introduced, using the Dutch program,
BLAISE. This has proved to save both time and effort as well
as reducing interviewer errors. In November 1992 and April
1993 answers to open questions about economic activity and
occupation continued to be handwritten but since November
1993 all responses have been entered immediately into a
computer by the interviewers.
Sample and response. The sampling frame for the ILFS
is drawn from the National Register comprising all Icelandic
and foreign citizens, 16-74 years of age, with domicile in
Iceland at the time of the survey. Icelandic persons who are
students outside Scandinavian countries are allowed to
retain their domicile in Iceland, whereas students in other
Nordic countries are required by Nordic agreements to
transfer their registration to the respective country. In order
to treat both groups of students in the same manner, all
persons who reside abroad for more than 6 months are
defined as ineligible in the final sampling frame. This
differs somewhat from intemational convention where all
students abroad are considered as part of the population,
irrespective of length of study.
The sample is a rotating panel sample of approximately
4,400 indi viduals selected by simple random method without
replacement from the sampling frame. The sample is divided
into four rotation groups of approx. 1,100 individuals, each
of which participates in four successive surveys. The sample
units are replaced into the sampling frame when four years
have elapsed from their first being selected.
In the first survey in April 1991 the size of the sample was
3,000, randomly divided into three rotation groups, the first
of which remained in the sample for four times, the second
participated three times and the third was dropped from the
sample after the second survey.
In the second to fifth surveys sampled indi viduals who had
previously refused to answer or were not able to participate
due to illness or disability were not contacted again. The
number of people contacted was thus smaller than the actual
sample. In November 1993 refusers (from earlier surveys)
were recontacted. Of these 37% refused again, 52% re-
sponded, 7% were not at home and 4% were unable to
answer due to illness or disability. In 1994 this effort of
refusal conversion was continued, albeit with less success.
The net response rate in the ILFS 1992-1994, i.e. the
number of respondents as percentage of all eligible sample
units, has been between 85 and 91%. The actual cooperation
rate, i.e. number of respondents as percentage of the net
working sample, was between 90 and 93% in the same period
(cf. Table 2.2).
3.1.3. Reliability
Errors in a sample survey fall roughly into two categories:
Sampling errors and non-sampling errors. This section deals
mostly with the errors that have affected the ILFS and is not
a comprehensive discussion of sample errors in general.
Sampling errors. Every sample incurs uncertainty due to
the method of sampling. Because of the random nature of
this uncertainty, it is possible to calculate the confidence
limits for the estimated numbers. These are shown in Table
2.3 for the 95% confidence level. If, e.g., the mean number